摘要: Stochastic filtering is the problem of estimating state a dynamic system after time passes and given partial observations. It fundamental to automatic tracking, planning, control real-world stochastic systems such as robots, programs, autonomous agents. This paper presents novel sampling-based algorithm. Its expected error smaller than sequential Monte Carlo sampling techniques fixed number samples, we prove show empirically. does so by deterministic action sequences then performing exact on those sequences. These results are promising for applications in natural language processing, robot control.
aaaipress.org 参考文章(22)
Thomas Dean, Keiji Kanazawa, Probabilistic temporal reasoning national conference on artificial intelligence. pp. 524- 528 ,(1988)
Michael L. Littman, Stephen M. Majercik, MAXPLAN: a new approach to probabilistic planning international conference on artificial intelligence planning systems. pp. 86- 93 ,(1998)
Subbarao Kambhampati, David E. Smith, Daniel Bryce, Sequential monte carlo in probabilistic planning reachability heuristics international conference on automated planning and scheduling. pp. 233- 242 ,(2006)
Paulo Mateus, António Pacheco, Javier Pinto, Amílcar Sernadas, Cristina Sernadas, Probabilistic Situation Calculus Annals of Mathematics and Artificial Intelligence. ,vol. 32, pp. 393- 431 ,(2001) , 10.1023/A:1016738205696
仁 大西, Pearl, J. (1988, second printing 1991). Probabilistic Reasoning in Intelligent Systems: Networks of Plausible Inference. Morgan-Kaufmann. Cognitive Studies: Bulletin of the Japanese Cognitive Science Society. ,vol. 1, ,(1994) , 10.11225/JCSS.1.2_100
Sequential Monte Carlo methods in practice Springer. ,(2001) , 10.1007/978-1-4757-3437-9
Michael I Jordan, Zoubin Ghahramani, Tommi S Jaakkola, Lawrence K Saul, None, An introduction to variational methods for graphical models Machine Learning. ,vol. 37, pp. 105- 161 ,(1999) , 10.1023/A:1007665907178
Uffe Kjærulff, A computational scheme for reasoning in dynamic probabilistic networks uncertainty in artificial intelligence. pp. 121- 129 ,(1992) , 10.1016/B978-1-4832-8287-9.50021-9
Andrew Y. Ng, Michael I. Jordan, PEGASUS: A policy search method for large MDPs and POMDPs uncertainty in artificial intelligence. pp. 406- 415 ,(2000)
Eyal Amir, Efficient approximation for triangulation of minimum treewidth uncertainty in artificial intelligence. pp. 7- 15 ,(2001)